University of Management & Technology

Department of Artificial Intelligence

School of Systems and Technology

Student ID:_____________________ Section: A2

Name:__________________________ Semester: Fall 2024
Program: BS AI Exam: Mid Term
Course Title: Machine Learning Academic Year: 2024-2025
Exam Date: 16 December, 2024 Time Allowed: 75 Mins
Resource Person: Dr. Ashfaq Ahmad Total Marks: 35

DO NOT OPEN THIS PAPER UNTIL TOLD TO DO SO

The instructions below must be followed strictly.
Failure to do so can result in serious grade loss
• Keep your eyes on your page.
• Switch off your mobile phones completely.
• Draw neat, labelled diagrams where necessary.
• No extra sheet will be given.
• Use blank spaces on the sheet for rough work.
• Use blue or black ball point or pen, please do not use lead pencils or any other
color.
• Lending/ Borrowing of calculators or anything is strictly prohibited.
• All Questions are self-explanatory. No queries will be entertained.

Questions Q#1 Q#2 Q#3 Q#4 TOTAL

Marks CLO 1 CLO 1 CLO 2 CLO 2
Obtained
Total
Marks
10 08 08 09 35
Question # 1 (CLO-1) (10 marks)
a) States the process to derive the best-fit line using the least squares method for the dataset with input
values X=[1,2,3,4,5] and corresponding output values Y=[2,4,6,8,10]. Calculate the slope and
intercept of the line. Show all steps clearly. (8 marks)

Solution:
∑𝑛𝑖=1(𝑥𝑖 − 𝑚𝑒𝑎𝑛 (𝑥)) × (𝑦𝑖 − 𝑚𝑒𝑎𝑛 (𝑦))
𝐵1 =
∑𝑛𝑖=1(𝑥𝑖 − 𝑚𝑒𝑎𝑛 (𝑥)2

𝐵0 = 𝑚𝑒𝑎𝑛(𝑦) − 𝐵1 𝑚𝑒𝑎𝑛(𝑥)
𝑛 = 5,

∑ 𝑋 = 1 + 2 + 3 + 4 + 5 = 15

∑ 𝑌 = 2 + 4 + 6 + 8 + 10 = 30

𝑋 − 𝑚𝑒𝑎𝑛(𝑋) = −14, −13, −12, −11, −10

𝑌 − 𝑚𝑒𝑎𝑛(𝑌) = −28, −26, −24, −22, −20
(𝑋 − 𝑚𝑒𝑎𝑛(𝑋))(𝑌 − 𝑚𝑒𝑎𝑛(𝑌) = 392, 338, 288, 242, 200

∑(𝑋 − 𝑚𝑒𝑎𝑛(𝑋))(𝑌 − 𝑚𝑒𝑎𝑛(𝑌) = 392 + 338, +288, +242, +200 = 1460

(𝑥𝑖 − 𝑚𝑒𝑎𝑛 (𝑥)2 = 196, 169, 144, 121, 100

∑(𝑥𝑖 − 𝑚𝑒𝑎𝑛 (𝑥)2 = 196 + 169 + 144 + 121 + 100 = 730

∑(𝑋 − 𝑚𝑒𝑎𝑛(𝑋))(𝑌 − 𝑚𝑒𝑎𝑛(𝑌) 1460

𝐵1 = = =2
∑(𝑥𝑖 − 𝑚𝑒𝑎𝑛 (𝑥)2 730
𝐵0 = 𝑚𝑒𝑎𝑛(𝑦) − 𝐵1 𝑚𝑒𝑎𝑛(𝑥)
𝐵0 = 30 − (2)(15)
𝐵0 = 30 − 30 = 0
Final Equation
𝑌 = 𝐵0 + 𝐵1 × 𝑋
𝑌 = 0 + 2𝑋
b) Identifies the missing lines in the Python code below to calculate the predictions based on the given
slope and intercept: (2 marks)
import numpy as np
X = np.array([1, 2, 3, 4, 5])
slope = ____ # Provide the calculated value
intercept = ____ # Provide the calculated value
Y_pred = ________ # Predict Y values
print(Y_pred)
Answer:
import numpy as np
X = np.array([1, 2, 3, 4, 5])
slope = 2 # Derived from the least squares method
intercept = 0 # Derived from the least squares method
Y_pred = slope * X + intercept
print(Y_pred)
𝑌_𝑝𝑟𝑒𝑑 = [ 2 4 6 8 10 ]

Question # 2 (CLO-1) (08 marks)

a) Calculate and state the probability of drawing an Ace from a standard deck of 52 playing cards. Then,
identify the probability of drawing an Ace given that the card drawn is a Spade. (4 marks)
Solution:
1. Probability of drawing an Ace (P(Ace)):
There are 4 Aces in a deck of 52 cards.
4 1
(𝑃(𝐴𝑐𝑒)) = = = 0.077
52 13
2. Probability of drawing an Ace given it's a Spade (P(Ace | Spade)):
There is 1 Ace among the 13 Spades.
𝟏
(𝑷(𝑨𝒄𝒆 | 𝑺𝒑𝒂𝒅𝒆)) = = 𝟎. 𝟎𝟕𝟕
𝟏𝟑
b) Using Stochastic Gradient Descent, calculate and state the updated weight and bias values after one
epoch. Given: (4 marks)
• Initial weight w=0, initial bias b=0
• Learning rate α=0.1
• Training data point: (2, 3).

Solution

Data Point (2, 3):

a. 𝑷𝒓𝒆𝒅𝒊𝒄𝒕𝒆𝒅 𝒗𝒂𝒍𝒖𝒆: 𝑦 ^ = 𝑤 ⋅ 𝑥 + 𝑏 = 0 × 2 + 0 = 0
b. 𝑳𝒐𝒔𝒔: 𝐿 = (𝑦 − 𝑦 ^ )2 = (3 − 0)2 = 9
c. Gradients
 𝜕𝐿 𝜕𝐿 𝜕𝑦 ^
= ×
𝜕𝑤 𝜕𝑦 ^ 𝜕𝑤
𝜕𝐿
= −2(𝟑 − 𝟎) × (𝟐) = −𝟏𝟐
𝜕𝑤
𝜕𝐿
= −2(𝟑 − 𝟎) = −6
𝜕𝑏
➢ Update parameters:
𝑤 = 𝑤 − 𝛼 ⋅ (−12)
𝑤 = 0 − 0.1(−12) = 0 + 1.2 = 1.2
𝑏 = 𝑏 − 𝛼 ⋅ (−3.6)
𝑏 = 0 − 0.1(−6) = 0 + 0.6 = 0.6
Question # 3 (CLO-2) (08 marks)
a) You are building a Naive Bayes classifier for a dataset with three classes 𝐶1 , 𝐶2 , 𝐶3 . For a new data
point X, you calculate the following probabilities:
𝑃(𝐶1|𝑋) = 0.4, 𝑃(𝐶2 |𝑋) = 0.35 𝑃(𝐶3 |𝑋) = 0.25
1. Uses the following probabilities for a Naive Bayes classifier with three classes to determine if
they are valid:
2. If 𝑃(𝑋|𝐶1 ) = 0.5, 𝑃(𝐶1 ) = 0.6 𝑎𝑛𝑑 𝑃(𝑋) = 0.75, 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒 𝑃(𝐶1 |𝑋) using Bayes’ theorem.
Does this match the given probability? (5 marks)

Solution:

1. Validity Check
𝑃(𝐶1 |𝑋) + 𝑃(𝐶2 |𝑋) + 𝑃(𝐶3 |𝑋) = 0.4 + 0.35 + 0.25 = 1
Probabilities are valid.
2. Use Bayes’ Theorem:
𝑃(𝑋|𝐶1 ) + 𝑃(𝐶1 )
𝑃(𝐶1|𝑋) =
𝑃 (𝑋 )
0.5 + 0.6
𝑃(𝐶1 |𝑋) = = 0.4
0.75
Matches the given probability.
b) Identifies the missing lines in the following code for computing the hypothesis using a sigmoid
function in logistic regression. Complete the code by filling in the missing lines for 𝑧, the sigmoid
function, and the hypothesis ℎ0 (𝑥) (3 marks)

import numpy as np
# Given weights and features
 theta = np.array([0.5, 1.2])
x = np.array([1, 2]) # Includes bias term
# Compute z
z = __________ # Fill in the equation for z
# Sigmoid function
def sigmoid(z):
return __________ # Fill in the equation for the sigmoid function
# Hypothesis
h_theta_x = __________ # Fill in the equation for the hypothesis
print("Hypothesis (h_theta_x):", h_theta_x)

Answer:

z = np.dot(theta, x)
return 1 / (1 + np.exp(-z))
h_theta_x = sigmoid(z)

Question # 4 (CLO-2) (09 marks)

Demonstrates the process of building a linear regression model using the hypothesis ℎ(𝑥) = 𝑤 ⋅ 𝑥 + 𝑏
to predict house prices (𝑦) based on the size of the house (𝑥). given a dataset with corresponding values:
House Size (𝑥) (in 𝑚2 ) Price (y) (in $)
50 200
100 400
150 600

1. Assuming the initial weights are 𝑤 = 0 𝑎𝑛𝑑 𝑏 = 0, calculate the predictions ℎ(𝑥) for the given data
points and the Mean Squared Error (MSE).
2. Explain what the high MSE at 𝑤 = 0 𝑎𝑛𝑑 𝑏 = 0 implies about the relationship between 𝑥 𝑎𝑛𝑑 𝑦.
3. Update the weight 𝑤 using one step of gradient descent with a learning rate 𝛼 = 0.01, assuming 𝑏 =
0 for simplicity.
4. How does optimizing 𝑤 while keeping 𝑏 = 0 affect the predictions and MSE? What does this suggest
about the importance of 𝑏 in the hypothesis function?

Solution: